| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 617.02 KB | Adobe PDF |
Advisor(s)
Abstract(s)
Como resultado da evoluĆ§Ć£o e inovaĆ§Ć£o tecnolĆ³gicas, a indĆŗstria da distribuiĆ§Ć£o electrĆ³nica de mĆŗsica tem tido um enorme crescimento. Desta forma, tarefas como a classificaĆ§Ć£o automĆ”tica de gĆ©neros musicais tornam-se um forte motivo para o incremento da investigaĆ§Ć£o na Ć”rea. O reconhecimento automĆ”tico de gĆ©neros musicais envolve tarefas como a extracĆ§Ć£o de caracterĆsticas das mĆŗsicas e o desenvolvimento de classificadores que utilizem essas caracterĆsticas. Neste estudo pretendeu-se, atravĆ©s de 3 problemas de classificaĆ§Ć£o independentes, classificar peƧas de mĆŗsica clĆ”ssica. Foi construĆdo um protĆ³tipo para um sistema real de classificaĆ§Ć£o, onde de um conjunto de mĆŗsicas nĆ£o catalogadas, foram automaticamente extraĆdos dez segmentos de seis segundos cada. Cada segmento musical foi classificado individualmente utilizando redes neuronais, tendo sido, para tal, extraĆdas 40 caracterĆsticas por segmento. Cada mĆŗsica foi classificada no gĆ©nero mais representado pelos seus segmentos.
As a result of recent technological innovations, there has been a tremendous growth in the Electronic Music Distribution industry. In this way, tasks such us automatic music genre classification address new and exciting research challenges. Automatic music genre recognition involves issues like feature extraction and development of classifiers using the obtained features. In this study we aim to classify classical music in subgenres, through three independent classification problems. Therefore, we extract 40 features for each one of the musical segments and we use neural nets as classifiers. Afterwards, due to the quality of the obtained results, a prototype system for automatic music classification of entire songs (not only segments) was built. We use 10 extracts for each song, uniformly distributed throughout the song. Each song is classified according to the most representative genre in all extracts.
As a result of recent technological innovations, there has been a tremendous growth in the Electronic Music Distribution industry. In this way, tasks such us automatic music genre classification address new and exciting research challenges. Automatic music genre recognition involves issues like feature extraction and development of classifiers using the obtained features. In this study we aim to classify classical music in subgenres, through three independent classification problems. Therefore, we extract 40 features for each one of the musical segments and we use neural nets as classifiers. Afterwards, due to the quality of the obtained results, a prototype system for automatic music classification of entire songs (not only segments) was built. We use 10 extracts for each song, uniformly distributed throughout the song. Each song is classified according to the most representative genre in all extracts.
Description
Keywords
ClassificaĆ§Ć£o de mĆŗsica Redes neuronais ExtracĆ§Ć£o de caracterĆsticas AnĆ”lise de sinais musicais Neural nets Music information retrieval Music classification Feature extraction
Pedagogical Context
Citation
MALHEIRO, Ricardo [et. al.] - Sistemas de classificaĆ§Ć£o musical com redes neuronais. GestĆ£o e Desenvolvimento. Viseu. ISSN 0872-556X. NĀŗ 12 (2004), p. 167-195.
Publisher
Universidade CatĆ³lica Portuguesa. Instituto UniversitĆ”rio de Desenvolvimento e PromoĆ§Ć£o Social